gregb/product-taxonomy-bench

Dataset Summary

product-taxonomy-bench is an anonymised benchmark dataset for predicting Shopify Product Taxonomy categories from Shopify product tags.

This dataset does not include raw product titles, raw tags, or product URLs. Tags are anonymised as tagNNNNNN.

Start Here

• Read this dataset card for the snapshot layout and field definitions.
• Open the benchmark notebook at notebooks/product_taxonomy_bench.ipynb. On the notebook page, use the Hub's Open in Colab button to run it interactively. The notebook defaults to the fixed paper snapshot.
• Use the snapshot folders according to your goal: paper/ for the canonical point-in-time paper snapshot, latest/ for the rolling benchmark, and first1000/ for a tiny sanity-check slice when present.

Configurations

Three configurations are provided:

Paper snapshot

• paper-2026-02-11T1915Z (created 2026-02-26T17:35:34.843938+11:00; 6,693 products, 2,542 tags, 363 taxonomies; as_of 2026-02-12T06:15:00+11:00)

Latest snapshot

• latest-2026-07-29T0722Z (created 2026-07-29T17:22:21.883007+10:00; 5,114 products, 1,866 tags, 347 taxonomies)

First 1000 snapshot

• first1000-2026-07-29T0722Z (created 2026-07-29T17:22:57.779736+10:00; 994 products, 11,126 tags, 309 taxonomies)

Data Fields

Each record corresponds to one product:

• product_id_hash: SHA-256 hash of a canonicalised product URL
• taxonomy_id: Shopify taxonomy GID
• taxonomy_path: Numeric hierarchy path (dot-separated) when available
• taxonomy_name: Human-readable hierarchy name
• cv_fold: 0–4 fold assignment (or null if missing)
• tag_features: list of {tag_id, in_title, title_part, title_position}

Tag semantics are not included; tag_id values are stable only within a snapshot.

Generation

Products were collected by fetching public Shopify product .json endpoints, then joined to the taxonomy label used by the cantbuymelove site. Tags are uppercased and substring-nested tags are filtered before anonymisation. Title overlap positions are computed by case-insensitive substring search and splitting titles on " - " to match the paper’s tag-battle logic. The paper snapshot is generated with a fixed as_of cutoff timestamp.

Citation

If you use this dataset, please cite the dataset release:

@misc{baker2026producttaxonomybench,
  author={Gregory D. Baker},
  title={product-taxonomy-bench: An anonymized benchmark for Shopify product taxonomy prediction from tags},
  year={2026},
  howpublished={\url{https://huggingface.co/datasets/gregb/product-taxonomy-bench}},
  note={Hugging Face dataset}
}